Mr. Sasvari,

Please, consider the content of the following publications [1-3].

[1] Stochastic Dynamic Mass Spectrometric Approach to Quantify Reserpine in Solution; Bojidarka Ivanova, Michael Spiteller

Analytical Chemistry Letters, 10 (2020) 703-721; Received 13 Oct 2020, Accepted 16 Dec 2020, Published online: 28 Jan 2021

Download citation https://doi.org/10.1080/22297928.2020.1865834

[2] Steroids, 164 (2020) 108750

Stochastic dynamic mass spectrometric quantification of steroids in mixture — Part II; Bojidarka Ivanova, Michael Spiteller

[3] B. Ivanova, M. Spiteller; EXPERIMENTAL MASS SPECTROMETRIC AND THEORETICAL TREATMENT OF THE EFFECT OF PROTONATION ON THE 3D MOLECULAR AND ELECTRONIC STRUCTURES OF LOW MOLECULAR WEIGHT ORGANICS AND METAL–ORGANICS OF SILVER(I) ION; pp. 1-182; in Book ''Protonation: Properties, Applications and Effect,'' A. Germogen (Ed.) NOVA Science Publishers, New York, (2019); ISBN: 978-1-5361-4886-2.

They deal with our own-authored (to me and my co-author's according to the shown authorship) innovative stochastic dynamic method and formulas for exact quantification of analytes by mass spectrometry.

Dear Peter Sasvari

The comparison you propose is very misleading. Briefly, we use (likely) trypsin to digest the proteins present in the sample. Therefore, depending on the lenght of your proteins you'll generate more or less peptides. Thus, peptide count is never a good option as it is highly biased for protein size. Must say that other factors also influence, e.g. PTMs decorating the proteins, which can affect enzyme accessibility to cleavage sites in a sample-specific maner.

A different approach would be to do spectral counting (based on PSMs; Peptide spectrum matches). This is a bit more accepted, but still not the best way to go. This type is based on MS2 evidence which can be affected by many factors making it not fully accurate either.

The most accepted practice in the field is to use MS1 (precursor ions) Intensity-based quantitation. I reccomend you to read this article about "label-free quantitation" (Article Accurate Proteome-wide Label-free Quantification by Delayed ...

In intensity-based methods, the intensity of precursor ions (a.k.a the peptide, MS1) that were identified based on spectral evidence (MS2) is summed/averaged/etc. to give a good estimate of protein abundance. Especially, if your data was acquired with MaxQuant (free available software), you should use LFQ Intensity, which applies a normalization across your samples to make them fairly comparable (for example, when one protein was identified with 3 peptides in sample A and 2 peptide in sample B, only the 2 common peptides will be used for quantification).

So, if you have intensity information, I would suggest to compare your samples using this parameter. If you don't have LFQ, but "abundance" from another software, you can check if you need to normalize so that you are doing a fair comparison.

Best regards,

Julia

Dear Julia Bauzá Martinez

Thank you for your answer, it was extremely helpful! I am new to the MaxQuant software so i am in the phase of trying to use it properly. I have run it with two of my recent samples and it calculated the LFQ intensities just fine if the protein could be found in both samples but in cases when only one of the samples had it, it calculated zero as LFQ intensity. For me these proteins can be important so i cannot omit them.

Will this anomaly be sorted out if i work with more (eg 6 or 8) samples together?

My other question is: how can i apply this to more samples? For example if i have 6 samples, 3 of which are control samples and the other 3 is my "sample of interest". If i want to know whether a protein is generally more abundant in my "sample of interest" compared to my control samples, what is the best way: taking the mean of the 3 controls ant the other 3 and compare these "mean LFQ intensities" or something else?

Best regards,

Peter

Dear Peter Sassari,

Glad to hear you started using MaxQuant, it's a great software accessible to everyone.

Good news is that this software comes with a sister software named Perseus, which you can use for data filtering, statistical analysis and representation. If you download the software, you can find useful manuals online. There is also tutorials in youtube: under the topic "MaxQuant summer school" you should find also a section of tutorials on Perseus software.

Anyway, replying to your question, you are dealing with Missing value. In general, we don't assume the proteins are absent, we prefer to assume that they are too low abundant to be detected. Therefore, one trick we normally use is to Log transform our data (log 2 is normally used), and then impute missing value according to the normal distribution of your sample. So, the proteins that had 0 LFQ will suddenly have a very low LFQ value, based on the distribution of the whole set of proteins in your sample. This will allow you to plot, generate a P-value, etc. However, I want to point out that normally I impute missing value only for proteins found in at least 2 out of 3 replicates in at least one of the groups.

Otherwise you can always present lists of presence/absence but representation is tricky.

And yeah, you can use what described above for your 3 control vs 3 treated, then log2 transform, impute missing value, T-test and yeah, compare the means of significant ones.

I hope you can manage to do this with Perseus and the available tutorials online. Good luck,

Julia